Not working on Nvidia Jetson Nano with Ubuntu 18.04 LTS

[erkarim85]

---

Required Info
Camera Model	{ D435i }
Operating System & Version	Linux (Ubuntu 18.04 LTS)
Platform	NVIDIA Jetson
Language	{python}

Issue Description

Hello,

I am trying to use the camera (D435i) on Jetson Nano using Python3 on Jupyter Notebook.
In order to do that, I needed to import pyrealsense2 as rs
However, I got a message saying it is not defined, so I tried pip install pyrealsense2. However, seems my Jetson Nano is not supported.
As a work around, I tried to build the source on the Jetson Nano using the terminal commands you provided.
After I went through everything, I still get the same error that it is not defined.
I tried copying the pyrealsense2.so and librealsense2.so to the same path of my code file, but it didn't work either.

I wonder if you can help me with my issue.

Thanks in advance.
Karim

[MartyG-RealSense]

Hi @erkarim85 Pip install packages cannot be used with devices with Arm processors such as Jetson and Raspberry Pi, as they are designed for x86 / x64 processors. There is a method for installing librealsense and Python that has worked consistently for Jetson Nano users though. The instructions are in the link below.

#6964 (comment)

If you are not using Python 3.6, change the two occurrences of '3.6' in the instructions to match the version number of Python 3 that you are using.

[erkarim85]

Great! Thank you very much @MartyG-RealSense. I will follow those instructions and try it and let you know.

[erkarim85]

Thanks again for your help @MartyG-RealSense.
I can now import pyrealsense2 without any errors.
However, now I am having another problem, I wanted to test it using the simple test code provided, but I got this error:

AttributeError: module 'pyrealsense2' has no attribute 'pipeline'

Any advice?

[erkarim85]

I tried to type:
dir(rs)
and I got:
['__doc__', '__loader__', '__name__', '__package__', '__path__', '__spec__']

[MartyG-RealSense]

There was a Jetson case involving AttributeError: module 'pyrealsense2' has no attribute 'pipeline' where the solution in that particular case was found to be to copy .so files to the same directory that the user's project script was in.

#7078 (comment)

This same approach of copying the .so files into the same directory as the user's project script also worked in a case with a Pi user:

https://support.intelrealsense.com/hc/en-us/community/posts/360048396854/comments/360012486393

[erkarim85]

So I tried that and it didn't work.
I tried one other thing, I uploaded cmake from 3.13 (the first link recommendation) to 3.18, and followed the same instructions for installation from the source.
I reach a weird situation:
I can see the package in the terminal, but I can't see it in Jupyter Notebook.

I checked from terminal for the file for this package and I put it in the ~/.bashrc file, and I copied this file into the same directory of my file used in jupyter. Still I can only see it on terminal but on Jupyter there is no 'pipeline' in rs module.

Any recommendations?

again, I appreciate your help.

[MartyG-RealSense]

There is a Start testing in Jupyter notebook guide in the librealsense documentation that may be of some help.

https://colab.research.google.com/drive/10YTLAf2i0R80-XX_6-1gTPtOMD0cRIx5

I note that the guide suggests installing pyrealsene2 using pip install pyrealsense2. This would not be applicable for your Jetson though, as Jetson uses an Arm processor and the pip packages for pyrealsense2 require an x86 or x64 processor like those in laptop / desktop computers. For Jetson Nano, the librealsense + pyrealsense2 method that I provided earlier in #6964 (comment) is recommended,

[MartyG-RealSense]

My research also found a case where someone was receiving the error DLL load failed while importing pyrealsense2: The specified module could not be found when using the command import pyrealsense2 as rs in Jupyter notebook.

In that particular case, they found that they were running their Jupyter notebook in a Python 3.7 virtual environment but Jupyter was picking up Python 3.8. Once they fixed this, their Jupyter notebook worked.

https://support.intelrealsense.com/hc/en-us/community/posts/360049911373-setting-up-D435i-to-work-in-Windows-10-and-Python-3-7

[erkarim85]

Thank you @MartyG-RealSense for your replies.
The example of using it won't work for me on jupyter yet, as I told you before, I can see the pyrealsense2 only on python terminal but Jupyter does not see the contents of the package.

I made sure before that I installed and used the package with the proper python version.
Jupyter Notebook is using Python 3.6.9 and I started python on the terminal using the same version.

I tried writing on terminal:

$ python3.6
>>> import pyrealsense2 as rs
>>> dir(rs)

This provided me with a lot of functions available in this package (around 20 lines on my terminal):

['BufData', 'STAEControl', 'STAFactor', 'STCensusRadius', 'STColorControl', 'STColorCorrection', 'STDepthControlGroup', 'STDepthTableControl', 'STHdad', 'STRauColorThresholdsControl', 'STRauSupportVectorControl', 'STRsm', 'STSloColorThresholdsControl', 'STSloPenaltyControl', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__spec__', '__version__', 'adjust_2D_point_to_boundary', 'align', 'auto_calibrated_device', 'calibrated_sensor', 'calibration_status', 'calibration_type', 'camera_info', 'color_sensor', 'colorizer', 'composite_frame', 'config', 'context', 'debug_protocol', 'decimation_filter', 'depth_frame', 'depth_huffman_decoder', 'depth_sensor', 'depth_stereo_sensor', 'device', 'device_calibration', 'device_list', 'disparity_frame', 'disparity_transform', 'distortion', 'dsm_params', 'event_information', 'extension', 'extrinsics', 'filter', 'filter_interface', 'firmware_log_message', 'firmware_log_parsed_message', 'firmware_logger', 'fisheye_sensor', 'format', 'frame', 'frame_metadata_value', 'frame_queue', 'frame_source', 'hdr_merge', 'hole_filling_filter', 'intrinsics', 'is_pixel_in_line', 'log', 'log_message', 'log_severity', 'log_to_callback', 'log_to_console', 'log_to_file', 'max_usable_range_sensor', 'motion_device_intrinsic', 'motion_frame', 'motion_sensor', 'motion_stream', 'motion_stream_profile', 'next_pixel_in_line', 'notification', 'notification_category', 'option', 'option_range', 'options', 'pipeline', 'pipeline_profile', 'pipeline_wrapper', 'playback', 'playback_status', 'pointcloud', 'points', 'pose', 'pose_frame', 'pose_sensor', 'pose_stream', 'pose_stream_profile', 'processing_block', 'quaternion', 'recorder', 'region_of_interest', 'roi_sensor', 'rs2_deproject_pixel_to_point', 'rs2_fov', 'rs2_project_color_pixel_to_depth_pixel', 'rs2_project_point_to_pixel', 'rs2_transform_point_to_point', 'rs400_advanced_mode', 'save_single_frameset', 'save_to_ply', 'sensor', 'sequence_id_filter', 'software_device', 'software_motion_frame', 'software_notification', 'software_pose_frame', 'software_sensor', 'software_video_frame', 'spatial_filter', 'stream', 'stream_profile', 'syncer', 'temporal_filter', 'terminal_parser', 'texture_coordinate', 'threshold_filter', 'timestamp_domain', 'tm2', 'units_transform', 'updatable', 'update_device', 'vector', 'vertex', 'video_frame', 'video_stream', 'video_stream_profile', 'wheel_odometer', 'yuy_decoder', 'zero_order_invalidation']

I tried on jupyter to write same code:

import pyrealsense2 as rs

dir(rs)

I only got the following line as output:
['__doc__', '__loader__', '__name__', '__package__', '__path__', '__spec__']

So, pipeline is not available in the package on the jupyter notebook. It is available on the terminal.

By the way, I tried to copy the *.so files to the same directory of my *.ipynb but it didn't fix it.

I appreciate if you have another advice...
Thanks again!

[erkarim85]

Another update:

This problem happens when I access the jupyter notebook by typing in the browser 192.xxx.xx.x:8888 or trying to access jupyter by typing this address on my computer connected to the Jetson nano.

However, I don't have this problem when I use the jetson nano as a computer and type in terminal jupyter lab.

Unfortunately, for my application, I need to use it in the headless mode.

[MartyG-RealSense]

Is the pyrealsense2 installation on the server and not on the local machine that you are using the browser address on?

The link below describes the setup of a headless Jupyter Notebook system where most of the data is on the server and its graphical GUI elements are on the local machine.

https://ljvmiranda921.github.io/notebook/2018/01/31/running-a-jupyter-notebook/

[MartyG-RealSense]

Hi @erkarim85 Do you require further assistance with this case, please? Thanks!

[erkarim85]

I still need a little bit more time. It didn’t work that way yet so I am trying to use it in a different way now. Best Regards, Karim H. Erian

…

On Dec 12, 2020, at 5:38 AM, MartyG-RealSense ***@***.***> wrote:  Hi @erkarim85 Do you require further assistance with this case, please? Thanks! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[MartyG-RealSense]

Adding a note to keep this case open for a further time period.

[MartyG-RealSense]

Hi @erkarim85 Do you have an update that you can provide us, please? Thanks!

[MartyG-RealSense]

Case closed due to no further comments received.

[erkarim85]

@MartyG-RealSense
Thank you for your help. The Camera is working now with the Jetson Nano.

[MartyG-RealSense]

Thanks very much @erkarim85 for the update - great news that you got it working!